基于YCbCr色彩空间和AdaBoost算法的人脸检测

Face Detection Based on YCbCr Color Space and Ada Boost Algorithm

针对彩色图像中复杂背景和可变光照的问题,是不利于人脸检测的,该文采用一种YCb Cr肤色空间的分割图像排除非肤色的区域,结合Ada Boost算法的人脸检测。首先对判断是否需要对图像进行光照补偿,然后转换肤色聚类比较好的YCb Cr肤色空间,用高斯模型进行建模,得到肤色似然图,再进行肤色分割,得到人脸候选区域,最后使用训练好的Ada Boost算法进行候选区域的人脸检测。仿真实验结果表明,本文方法对复杂背景下彩色图像的人脸的检测率比较高。

For the problem of complex background and variable illumination in color images,it is not conducive to face detection.In this paper,a YCbCr skin color segmentation image is used to exclude non skin regions,combined with AdaBoost algorithm for face detection.First,to determine whether the need for image illumination compensation,and then convert color clustering better YCbCr color space modeling using Gauss model,color likelihood map,then the skin color segmentation,face detection and face candidate regions can be obtained,finally using the trained AdaBoost algorithm for candidate regions.The simulation results show that the detection rate of color images in complex background is high.